Since many epidemiologic problems cannot be solved by standard techniques, new methods are required to extract more complete answers from research data. The objective of this project is to use mathematical theory and computer simulations to develop and evaluate statistical methods appropriate to data arising in epidemiologic research, and to carry out the statistical programming needed to make these methods easily available to other researchers. This may include evaluating outside computer software, using standard programs in novel ways, and writing special purpose programs. Further study will continue in then use of influence statistics and regression diagnostics, in particular using the SAS procedures for regression and generalized linear models. Methods appropriate to categorical data and contingency tables will also be given special attention. Useful techniques will be presented in seminars and publications in statistical journals, as well as applied to data analysis within the Branch.